Maintaining a population of stable synaptic connections is probably of critical importance for the preservation of memories and functional circuitry, but the molecular dynamics that underlie synapse stabilization is poorly understood. Here, we use simultaneous time-lapse imaging of post synaptic density-95 (PSD-95) and Ca2+/calmodulin-dependent protein kinase II (CaMKII) to investigate the dynamics of protein composition at axodendritic (AD) contacts. Our data reveal that this composition is highly dynamic, with both proteins moving into and out of the same synapse independently, so that synapses cycle rapidly between states in which they are enriched for none, one or both proteins. We assessed how PSD-95 and CaMKII interact at stable and transient AD sites and found that both phospho-CaMKII and PSD-95 are present more often at stable than labile contacts. Finally, we found that synaptic contacts are more stable in older neurons, and this process can be mimicked in younger neurons by overexpression of PSD-95. Taken together, these data show that synaptic protein composition is highly variable over a time-scale of hours, and that PSD-95 is probably a key synaptic protein that promotes synapse stability.
Circuit maturation and experience-dependent plasticity are thought to involve the fine-tuning of synaptic connectivity within specific microcircuits. Presumably neurons must form and then maintain a population of stable contacts with specific synaptic partners in order to ensure proper circuit function, but the molecular mechanisms that underlie this process of synapse stabilization are still poorly understood [1,2]. Further, the ability of a synapse to remain stable is complicated by the recent demonstration that many synaptic proteins are highly dynamic and turn over in both activity-dependent and independent manners [3–7]. Here, we used time-lapse imaging of synapse formation in cultured neocortical neurons to examine the dynamics of association of Ca2+/calmodulin-dependent protein kinase II (CaMKII) and postsynaptic density-95 (PSD-95) with sites of axon–dendrite (AD) contact, in order to assess how the localization and amount of these proteins correlates with changes in contact turnover.
Both CaMKII and PSD-95 are localized at excitatory synapses, and have been implicated in promoting synapse stability. CaMKII expression is important for synapse formation and maturation [8–12], and previous work from our laboratory demonstrated that overexpression of constitutively active CaMKII enhances connectivity between active presynaptic partners by increasing new contact formation [13,14]. The scaffold protein PSD-95 is highly abundant at , and clusters very early at [16–18] excitatory synapses. Further, PSD-95 is more strongly associated with stable than transient spines , and is thought to accelerate synaptic maturation [3,20–22] through coordinated regulation of postsynaptic AMPA-type glutamate receptors (GluARs) [23–25] and neuroligins [26,27]. Synaptic PSD-95 levels increase during development and with time in vitro [28,29], over a similar developmental window there is a reduction in synapse turnover , and knockdown of PSD-95 increases spine turnover . These studies raise the possibility that CaMKII and/or PSD-95 might play critical roles in stabilizing synaptic contacts.
In the following set of experiments, we investigated the role of CaMKII and PSD-95 in synapse turnover and stability. We used time-lapse imaging of connected pairs of cultured neocortical pyramidal neurons to examine the dynamics of fluorescently tagged CaMKII and PSD-95 at sites of AD contact. One advantage of this approach is that we can monitor pre- and postsynaptic structures simultaneously, something difficult to do in most slice culture or in vivo models. We found that PSD-95 and CaMKII dynamically associate and dissociate from sites of both transient and stable AD contacts, but that these synaptic proteins spend more time at stable contacts. Additionally, stable AD contacts show an increase in the fraction of CaMKII in an active, phosphorylated state. Finally, we studied the effects of overexpressed PSD-95 on the stability of putative synaptic contacts, as assessed by monitoring the dynamics of postsynaptic GluA2 clusters that colocalize with a marker of presynaptic function (FM4-64). We found that overexpression of PSD-95 reduced the turnover rates of pre- and postsynaptic structures, thus promoting the stabilization of putative synaptic contacts. These data suggest that recruitment of PSD-95 into the postsynaptic compartment is one factor that contributes to the stabilization of synaptic contacts.
2. Material and methods
(a) Transfection of cortical cultures
Cultures were prepared as described previously [13,14] from P1 to P3 Long-Evans rat pups, and were transfected between 6 and 9 days in vitro (DIV) using the Lipofectamine 2000 (Invitrogen) reagent according to the manufacturer's instructions. Twelve to 16 h prior to the experiment, cultures were co-transfected with green fluorescent protein (GFP) tagged CaMKII (CaMKII-GFP)  and Discosoma red-tagged PSD-95 (PSD-95-DsRed) , while a separate population of neurons was transfected with cyan fluorescent protein (CFP) in order to visualize axons. In a second set of experiments cultures were transfected at either DIV 6–7 or DIV 14–15 with enhanced yellow fluorescent protein-tagged GluA2 (GluA2-EYFP) , alone or in combination with PSD-95-CFP (made by replacing DsRed with CFP in the vector described in  for 48 h prior to FM4-64 labelling).
(i) Time-lapse Imaging
Time-lapse imaging was performed as described previously [14,33]. Pyramidal neurons that expressed both CaMKII-GFP and PSD-95-DsRed and were contacted by a CFP-expressing axon from a neighbouring pyramidal neuron were identified. For FM4-64 experiments the dye was added for 2 min in a high K+ HEPES-buffered solution in the presence of 50 µM d-2-amino-5-phosphonovaleric acid, 20 µM 6-cyano-7-nitroquinoxaline-2,3-dione and 10 µM tetrodotoxin, then washed; at the end of the imaging period, cultures were re-labelled with FM4-64 to label any newly formed presynaptic puncta. Images were acquired every 20 min (2 µM Z stack, 0.2 µM step size) on an Olympus IX-70 using a 60× 1.25 na oil-immersion objective and a cooled CCD camera (Orca R2, Hamamatsu) controlled by Volocity software (Improvision), as described in ; pixel size was 0.11 µm. AD contacts refer to the physical contact between the fluorescently labelled AD. Puncta contacts were defined as an AD contact with a punctate fluorescent signal more than 25% above the local dendritic background; puncta were considered colocalized if they were within 1 pixel of each other. AD contacts could contain zero, one or multiple CaMKII-GFP and/or PSD-95-DsRed puncta. Stable contacts were defined as those that persisted throughout the 3 h experiment. Average contact lifetime refers to the average lifetime of all contacts, including transient and stable contacts, over the course of the 3 h experiment. Contact gain and loss were quantified as described .
After transfection with CaMKII-GFP and PSD-95-DsRed, cultures were fixed and stained as described previously , using an epifluorescence microscope equipped with a 60× oil-immersion objective (na 1.25) and the appropriate filters. Primary antibodies were anti-VGlut1 (guinea pig, 1 : 500, Synaptic Systems) or anti-phospho-T286 (mouse, 1 : 100, Abcam). The total area of the puncta or the total AD contact area was used as a mask and overlaid onto the phospho-T286 signal to allow quantification of signal intensity at these sites. Gained and lost contacts consisted of two types: those where a punctum was formed de novo or disappeared, and those where a punctum moved towards or away from an axon. Lost contacts where the punctum disappeared were excluded from this analysis since there was no punctum area to use as a mask. Gained and lost contacts had similar phospho-CaMKII levels so were combined into one category, labelled ‘transient’.
All data are reported as mean ± s.e.m. with the n referring to the number of neurons or puncta as indicated. Statistical analyses were made with unpaired two-tailed Students t-tests, or for multiple comparisons single factor ANOVAs followed by corrected t-tests using a Bonferroni correction for multiple comparisons.
A number of studies have shown that CaMKII and PSD-95 puncta are dynamic [3,4,7,17,34] but how these dynamics relate to the formation, loss and stabilization of synaptic contacts where both pre- and postsynaptic structures can be followed simultaneously has not been studied. To begin to understand how PSD-95 and CaMKII are distributed at stable and labile AD connections, we first assessed their static localization rates to presynaptic sites in fixed cultures, using either sites of AD contact (figure 1a) or the excitatory presynaptic marker VGlut1. We showed previously that these AD contacts have approximately 80% colocalization with Synapsin, GluA2 and FM dyes , suggesting that the majority of these sites represent functional synaptic contacts. Consistent with this previous report, here we found a high degree of colocalization between individual markers and either VGlut1 (figure 1b) or AD contacts (figure 1c). Interestingly, significantly fewer of these putative synaptic sites had PSD-95 and CaMKII together than PSD-95 alone (figure 1b,c). These data indicate that although localization rates of individual synaptic markers at synaptic sites are fairly high, only a fraction (about 30%) of these sites have both PSD-95 and CaMKII simultaneously. This suggests that the dynamics of these synaptic proteins are to some degree independent, so that the complement of proteins at existing AD contacts detected at any one time point will be quite variable.
To directly examine the dynamics of PSD-95 and CaMKII at AD contacts, we performed time-lapse imaging on cultured rat visual cortical neurons sparsely transfected with PSD-95-DsRed and CaMKII-GFP, that were contacted by an axon expressing CFP from a neighbouring pyramidal neuron. This allowed us to follow the formation and loss of AD contacts, as well as the movement of PSD-95 and CaMKII puncta, into and out of these sites of contact. We first looked at the stability of AD contacts and the stability of PSD-95 or CaMKII puncta at these contact sites (referred to as puncta contacts); we found that AD contacts are significantly more stable, and have longer contact lifetimes, than PSD-95 or CaMKII puncta contacts (figure 1d,e). Further, PSD-95 puncta contacts were more stable and had longer lifetimes than CaMKII puncta contacts (figure 1d,e), consistent with our static colocalization data (figure 1b).
To examine turnover of AD and puncta contacts, we next quantified the rates of contact gain and loss. Consistent with our previous work , we found that rates of gain and loss were well balanced for each contact type, suggesting that overall synapse number and composition were relatively stable. When we compared across contact type, we saw that AD contacts have significantly lower turnover than either PSD-95 or CaMKII puncta, which is consistent with their higher stability and longer contact lifetimes (figure 1f). Thus, physical AD contacts are much more stable than their constituent postsynaptic proteins, which can move in and out of existing AD contacts.
Next we asked whether the behaviour of PSD-95 and CaMKII puncta were different at stable and transient AD contacts. Figure 2a shows two examples of AD contact formation and the accompanying protein dynamics. The top panel illustrates an axon making contact with a preexisting dendritic site at which PSD-95 and CaMKII are colocalized. The bottom panel illustrates a dendritic PSD-95 punctum that moves towards the axon and makes a contact that persists until the end of the experiment. In figure 2b, two examples of stable AD contacts are illustrated: the top panel illustrates a stable AD contact that remains colocalized with both PSD-95 and CaMKII throughout the imaging session, and the bottom panel illustrates a contact that transiently accumulates CaMKII but not PSD-95 (note that the nearby PSD-95 punctum does not colocalize with the site of contact or CaMKII accumulation). We observed a great variety of such dynamics at both stable and transient AD contacts; to quantify this behaviour we calculated the percentage of transient and stable AD contacts containing one or both synaptic proteins at any point during the experiment, as well as the percentage that did not significantly accumulate either. Strikingly, stable contacts were much more likely to have both synaptic proteins present at some point during the experiment (figure 2c). To quantify this further, we calculated the percentage of time that each puncta type spends at stable and transient AD sites (figure 2d). PSD-95 spent significantly more time at stable than at transient AD contacts, and also spent significantly more time at stable AD contacts than either CaMKII or colocalized PSD-95 and CaMKII puncta. Though there was a trend for more CaMKII and colocalized PSD-95 + CaMKII at stable contacts, this was not statistically significant. Thus, PSD-95 spends significantly more time at stable AD sites than at transient sites. This raises the possibility that PSD-95 plays a role in the stabilization of those AD sites.
Previously, we found that activated CaMKII enhances connectivity between pairs of neurons by increasing the rate of new contact formation [13,14]. Thus, while there was no significant difference in dwell-time of CaMKII at stable or transient connections (figure 2d), the important variable might be the activation state of CaMKII rather than absolute protein levels. To address this, cultures were fixed after time-lapse imaging of CaMKII and PSD-95 and probed using an antibody specific for phospho-T286 (figure 3a). The area that encompassed the PSD-95 or CaMKII punctum was used as a mask and overlaid onto the phospho-T286 staining, to allow us to correlate the intensity of phospho-T286 with the behaviour of either the CaMKII or PSD-95 punctum. For the AD contacts, a mask was made that encompassed the total area of contact and this was dilated by 1 pixel in order to account for puncta that were partially overlapping the site of AD contact. Interestingly, the ratio of activated to total CaMKII was lower at transient than at stable AD contacts (figure 3b). By contrast, this ratio was not correlated with the behaviour of PSD-95 or CaMKII puncta at AD contacts (figure 3b). Unfortunately, this post hoc approach did not give us access to the activation state of CaMKII just prior to contact gain or loss, so our data cannot rule out a transient enhancement of activated CaMKII just prior to formation of new contacts. These data show that, under our experimental conditions, there is more activated CaMKII at stable AD contacts.
We have shown so far that stable AD contacts have a higher ratio of active to inactive CaMKII and an enhanced accumulation of PSD-95. The synaptic localization of PSD-95 increases developmentally and with time in vitro [28,29], and there is ample evidence that PSD-95 overexpression can enhance the number of synapses in immature neurons [20,21,23,35], suggesting an important role for PSD-95 in some aspect of excitatory synapse formation. However, it is not known whether PSD-95 exerts its effects by enhancing the rate of synapse formation, or by enhancing the stability of newly formed contacts. We thus wished to assess whether overexpression of synaptic PSD-95 affects the turnover and stabilization of synaptic contacts.
In order to simultaneously follow a large number of presynaptic and postsynaptic sites on individual neurons with or without overexpressed PSD-95, we used live imaging of GluA2-EYFP to image synaptic receptor clusters , and FM4-64 dye loading to follow presynaptic contacts onto GluA2-EYFP expressing neurons; at the end of each imaging session a second round of FM loading was performed to identify newly formed presynaptic sites. As there is considerable surface dendritic GluA2, we were also able to follow dendritic structures with this technique (figure 4a). We have extensively characterized the effects of GluA2 expression on synaptic function and have found no effects on synaptic transmission or on synapse formation [32,33], so this method allows us to simultaneously image pre- and postsynaptic structures with minimal impact on synapse formation. Using this method, we were able to track sites where pre- and postsynaptic structures were stable throughout the experiment, and sites where pre- and postsynaptic contacts were either gained or lost (figures 4a and 5a).
In order to test the role of PSD-95 in promoting stabilization of AD contacts, we first quantified overall GluA2 dynamics regardless of their presynaptic association under baseline conditions and compared that to conditions where PSD-95 was overexpressed. To assess whether the effects of PSD-95 change as synapses mature, these dynamics were examined at two ages: DIV 7 and DIV 14. Between DIV 7 and DIV 14 GluA2 puncta became more stable; further, PSD-95 overexpression in young but not older neurons mimicked this increase in stability (figure 4b). Similarly, the average lifetime of GluA2 puncta was significantly longer at DIV 14 than at DIV 7, and this was also mimicked by overexpression of PSD-95 in young neurons (figure 4c). A similar effect of PSD-95 was observed when we quantified the rates of loss and gain for colocalized GluA2 and FM4-64 puncta, which are putative sites of synaptic contact (figure 5b). These rates were higher at DIV 7 than at DIV 14. Further, PSD-95 overexpression reduced turnover at DIV 7 but not at DIV 14. Interestingly, PDS-95 overexpression did not affect the ratio of contact gain to loss, as this ratio remained close to unity across ages and PSD-95 expression levels (figure 5b). These data show that PSD-95 overexpression in young neurons enhances the stability of putative synaptic contacts, while having little impact on older synapses that express more endogenous PSD-95 .
In this study we used simultaneous time-lapse imaging of two important synaptic proteins, PSD-95 and CaMKII, to probe the dynamics of protein composition at young synapses. Our data reveal that this composition is highly dynamic, with both proteins moving into and out of the same synapse independently, so that synapses cycle rapidly between states in which they have none, one, or both proteins present. Despite this dynamism, both PSD-95 and CaMKII spent more time at stable than at labile synaptic contacts. Further, we found that stable contacts have a higher ratio of activated to total CaMKII than do labile contacts. Finally, we found that overexpression of PSD-95 in young neurons (where expression levels are low) was able to reduce contact turnover and stabilize pre- and postsynaptic structures. In contrast, in older neurons with more stable contacts PSD-95 overexpression had no impact on synapse stability. Taken together, these data suggest that PSD-95 promotes synapse maturation and stabilization.
Consistent with previous reports [3–5,7], we found that PSD-95 and CaMKII are quite dynamic at AD contacts, moving into and out of these sites over the course of our 3 h experiment. Less than 20% of PSD-95 and CaMKII puncta were stable over this period of time, and the average lifetime of contacts was about 100 min for PSD-95 and 40 min for CaMKII; as a consequence putative synaptic sites cycle quickly between expressing one or the other or both of these proteins. These striking dynamics suggests that synapses have a highly variable protein composition over time, though the purpose and regulation of this variability are not clear. The majority of stable contacts have at least a transient expression of both PSD-95 and CaMKII such that PSD-95 spends significantly more time at stable AD contacts. This could be because the prolonged presence of PSD-95 helps to stabilize AD contacts, or because stable AD contacts are better able to accumulate PSD-95. The former interpretation is favoured by the observation that overexpression of PSD-95 increased the fraction of stable contacts.
PSD-95 binds indirectly to α-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid receptor (AMPAR), regulates synaptic strength, and has been proposed to serve as a ‘slot’ protein that directly determines synaptic AMPAR content [36–39]. Manipulations of PSD-95 can dramatically alter evoked synaptic transmission [23,25,37,40–43], but we found previously that PSD-95 overexpression or knockdown in this same culture system did not significantly increase AMPA quantal amplitude , in keeping with other reports of only modest effects [5,23,37,40,41,44]. The modest impact of PSD-95 on quantal amplitude, coupled with the substantial impact on synapse stability shown here and suggested previously , suggest that the effects of PSD-95 overexpression on evoked transmission are primarily due to an enhancement of the number of stable, functional synapses.
Interestingly, the effects of PSD-95 on synapse stabilization appear to be developmentally regulated, as PSD-95 overexpression stabilized young but not older synapses. This is consistent with our previous finding that PSD-95 knockdown reduces metabotropic excitatory postsynaptic current frequency in old but not young neurons . Between DIV 7 and DIV 14 the levels of synaptic PSD-95 increase substantially [28,29], and it has been shown that postsynaptic densities share and compete for PSD-95 [5,7]. As the acquisition of PSD-95 is probably regulated by changes in activity and experience [5,45,46] when PSD-95 levels are low early in development, competition for PSD-95 could help determine which synapses are maintained and which are lost. Later in development PSD-95 levels are high enough that they are probably no longer limiting, and so this form of competition may be restricted to very young synapses. This could explain why overexpression of PSD-95 stabilizes young synaptic contacts, but has no effect in older neurons.
This work was supported by NIH grant NS036853.
We thank L. Wang for technical assistance, and R. Nicoll, D. Bredt and M. N. Waxham for providing DNA constructs.
One contribution of 35 to a Discussion Meeting Issue ‘Synaptic plasticity in health and disease’.
- © 2013 The Author(s) Published by the Royal Society. All rights reserved.